Abstract
The incidence of breast cancer in the United States has risen by 1 percent per year since 1940 (1), and changes in the prevalence of established risk factors do not completely explain this increase. Scientists have hypothesized that the increase in breast cancer rates may be related to exposure to extremely low frequency (50-60 Hz) electric and magnetic fields (EMFs) produced from electric power facilities and household appliances (2). Beginning in the 1930s, when many electric appliances were introduced into the home, the use of electric power in the United States has grown steadily, approximately 7-13 percent per year through the 1970s (3). Furthermore, there is experimental evidence that exposure to low-level EMFs suppresses the secretion of melatonin, a hormone that might protect against breast cancer (4).
The majority of epidemiologic studies addressing EMFs and breast cancer have focused on exposure in occupational cohorts (5). A few studies have explored the risk associated with overhead power lines (5), and two groups have studied use of electric blankets (6-8). Electric blanket use represents a substantial source of close, whole-body exposure to EMFs. The contribution of an electric blanket turned on throughout the night to a user’s total exposure to EMFs has been estimated to be from two to four times those of ambient background residential levels (9, 10). Furthermore, electric blankets are usually used at night when the natural melatonin peak occurs, suggesting a specific mechanism for an influence on breast cancer risk. Electric heaters in waterbeds produce similar EMFs; however, because the source is farther from the user under the water-filled mattress, the exposure is considered unimportant (11). Case-control studies of 290 premenopausal (7) and 382 postmenopausal (6) cases observed a nonsignificant 40 percent increased risk with exposure to electric blankets turned on throughout the night, while a large case-control study of 2,199 cases under the age of 55 years found no association with similar exposure (8).
We examined the relation of electric blanket and waterbed use to breast cancer risk prospectively (exposure was assessed prior to disease diagnosis) and retrospectively (exposure was ascertained after breast cancer had occurred) in the Nurses’ Health Study. We examined the timing and duration of use and assessed the risk of invasive breast cancer and subsets defined by hormone receptor status. Because previous research has stratified by menopausal status and because the influence of alterations in circulating estrogen might differ by menopausal status, we also examined pre- and postmenopausal breast cancer separately.
MATERIALS AND METHODS
Study population
The Nurses’ Health Study is an ongoing prospective cohort study established in 1976 when 121,700 female registered nurses completed a mailed questionnaire that included items about risk factors for breast cancer and other diseases. At enrollment, the participants were between the ages of 30 and 55 years and resided in 11 states (California, Connecticut, Florida, Maryland, Massachusetts, Michigan, New Jersey, New York, Ohio, Pennsylvania, and Texas). Every 2 years, participants completed follow-up questionnaires to update information on risk factors for breast cancer and to report the occurrence of breast cancer and other illnesses. A series of questions on electric blanket and heated waterbed use were included on the first three mailings of the 1992 questionnaire. Of the 104,064 participants followed through 1992, 89,376 responded to one of these mailings. The subjects eligible for this analysis were the 87,497 women (98 percent of the responders) who answered the overall electric blanket and waterbed question: “Have you ever regularly used an electric blanket or a heated waterbed?”
Assessment of outcome
Diagnoses of breast cancer were reported on the biennial follow-up questionnaire. Nonrespondents were contacted by telephone, and deaths were identified through next of kin or searches of the National Death Index. For each case of breast cancer reported, permission to obtain medical records and pathology reports to confirm the diagnosis was requested. Because the accuracy of self-reported breast cancer was extremely high (12), the small number of cases for whom pathology reports were not obtained were included in this analysis. Estrogen receptor status was extracted from the medical records.
Assessment of exposure
In 1992, participants answered separate questions on lifetime use of either an electric blanket or waterbed. Information was obtained on the number of years, in prespecified categories, participants had regularly used an electric blanket kept running overnight or a heated waterbed during three time periods: 1989-1991 (0, 1, 2, or 3 years), 1976-1988 (0, 1-3, 4-7, or 8-13 years), and before 1976 (0, 1-6, 7-14, or >=15 years). A continuous variable for years of electric blanket use over all time periods was created by summing the middle value of the intervals for each category of use. If information was available for at least one time period, but not for all three, then we assumed that there was no use during the missing periods and set those periods to zero for the purposes of calculating the continuous variable. For time period-specific analyses we did not make any assumptions about missing data and left those responses as missing. In addition, we asked at what age each participant first regularly used either appliance (0-9, 10-19, 20-44, or ≥45 years old). Separate questions were asked for electric blankets and waterbeds, and separate analyses were conducted for each exposure.
Assessment of covariates
Information on known and suspected risk factors for breast cancer was obtained from the biennial questionnaires. The following risk factors were included in the multivariate models: age, menopausal status, age at menopause, age at menarche, parity, age at first full-term pregnancy, use of oral contraceptives, use and duration of use of postmenopausal hormone therapy, history of breast cancer in a mother or sister, history of benign breast disease, height, current body mass index (weight (kg)/height (m)2), body mass index at age 18 years, region of residence, and alcohol consumption. A woman was classified as postmenopausal if she returned a questionnaire on which she reported either natural menopause or hysterectomy with bilateral oophorectomy. Women reporting hysterectomy without bilateral oophorectomy were considered to be postmenopausal at the age when natural menopause had occurred in 90 percent of the cohort (54 years for current cigarette smokers and 56 years for nonsmokers) and premenopausal if younger than 46 years of age for smokers and greater than 48 years for nonsmokers. Otherwise, they were considered to be of uncertain menopausal status.
Data analysis
This analysis consisted of two parts: a prospective analysis from 1992 to 1996 in which exposure to electric blankets and waterbed use was assessed prior to the occurrence of breast cancer, and a retrospective analysis for the years 1976-1992 in which exposure was ascertained after diagnosis. The prospective analysis was restricted to women who had not been diagnosed with cancer, with the exception of non-melanoma skin cancer, prior to 1992 (n = 78,614) and answered the 1992 questionnaire on exposure to EMFs. The retrospective analysis included those women who answered the questions on EMFs and were cancer free in 1976, the start of the Nurses’ Health Study (n = 85,474). The person-months of follow-up were accumulated for each risk factor, and risk factor status was updated when new information was available; for most variables this was every 2 years. Women accumulated person-time of exposure until the date of diagnosis of breast cancer, diagnosis of other cancer, or death, whichever occurred first. Relative risks were calculated as the rate of disease among the exposed divided by the rate of disease among the nonexposed. Mantel-Haenszel summary relative risks were calculated, adjusting for age in 5-year categories. To control simultaneously for potential confounding factors, we fit proportional hazard models (13) using a pooled logistic regression approach (14, 15) with indicator variables for each category of exposure, covariate, and 2-year interval of calendar time. Ninety-five percent confidence intervals were calculated for each relative risk. In the prospective analysis, four definitions of exposure were assessed in separate models, all using “never use” as the reference category: 1) any use; 2) any use in the most recent time period, 1988-1991; 3) any use in the most distant time period, before 1976; and 4) categories of total years of use. Any electric blanket use, electric blanket use before 1976 (the beginning of disease follow-up), and categories of years of use before 1976 compared with never use were evaluated in separate retrospective models. Analyses were performed for all women combined and stratified by menopausal status. In addition, all analyses were repeated restricting the case definition to those women with estrogen receptor-positive breast cancer.
Exposure reproducibility substudy
To assess the reproducibility of the questions on EMFs on the 1992 questionnaire, we mailed a supplemental questionnaire to a stratified random sample of 200 women who reported some use of electric blankets, 200 women who reported some use of a heated waterbed, and 100 women who answered “no” to both questions. The response rate was 85 percent after three mailings. The supplemental questionnaire queried, in a grid format, specific use of an electric blanket kept running throughout the night and of a heated waterbed as the primary bed for each individual year between 1976 and 1991. Use before 1976 was assessed by a question identical to the one included on the 1992 questionnaire. Information on the number of months in each year of waterbed use and on the number of months in each year and nights per month of electric blanket use was also obtained. This more detailed information was used to determine the reproducibility of the original answers and to better characterize the average usage profile for electric blankets and heated waterbeds.
RESULTS
In 1992, 48 percent of the cohort reported regular use of a waterbed or an electric blanket throughout the night during sometime in their lives. Forty-two percent of the cohort had used an electric blanket (n = 36,805), 7 percent had used a waterbed (n = 6,487), and 4 percent had used both (n = 3,360) (9 percent of the electric blanket users and 52 percent of the waterbed users). At baseline, 1992, electric blanket users and never users were similar in age (mean age for both groups was 59 years with a standard deviation of 7 years) and had similar age-adjusted prevalences of most established breast cancer risk factors (table 1). Electric blanket users were slightly more likely to have used oral contraceptives and postmenopausal hormones for a long period of time, and they were slightly more likely to have been diagnosed with benign breast disease. The percentage living in California was greater among users (17 percent) than among never users (9 percent). The comparison of the distribution of risk factors among ever waterbed users and never users was very similar; waterbed users were also slightly more likely to use oral contraceptives and postmenopausal hormones than were never users and to live in California (data not shown). Ever waterbed users were slightly younger (mean age, 55 years; standard deviation, 7) than were never users (mean age, 59 years; standard deviation, 7). Since only 3 percent of the electric blanket users reported that they had first used one earlier than age 19, the first exposure during childhood or puberty could not adequately be assessed.
Reproducibility of electric blanket information
Among the respondents in the reproducibility substudy, comparisons between the categorical responses on the original 1992 questionnaire and the more detailed responses on the 1994 supplemental questionnaire indicated that reports of no use during a given time period were reasonably reproducible. In the supplemental questionnaire, we arbitrarily defined regular use as use of an electric blanket throughout the night on at least seven occasions during the course of a year. On average, these women used an electric blanket at least weekly during 6 months of the year for 6 hours a night. The agreement of a response of “never use,” calculated as the probability that a person reported in 1992 no electric blanket use during any of the three time periods (before 1976, 1976-1988, and 1988-1991) given that she reported no use on the supplemental questionnaire, was 91 percent. This percentage was highest for the period before 1976 (96 percent) and lowest during the 1976-1988 time period (83 percent). Agreement of response of “any use” during a time period or over all three time periods (the probability of reporting use in 1992 given that she had also reported use on the supplemental questionnaire) was not as consistently high, ranging from 64 percent in the earliest time period to 92 percent in the most recent period (table 2). For categories of duration of use within a time period, the percentage identically classified or in an adjacent category is also presented in table 2, along with the Spearman correlation coefficients comparing continuous years of electric blanket use estimated from the supplemental and original questionnaires. The Spearman correlation coefficient for years of total use across all three time periods was 0.75 (p = 0.0001).
Breast cancer incidence
In the prospective analysis, 954 incident cases of invasive breast cancer occurred between 1992 and 1996 among 78,614 nurses. The age-adjusted relative risks for any use of an electric blanket and for use at least 16 years prior to diagnosis (before 1976) were elevated (table 3). Controlling for known breast cancer risk factors attenuated these results, and they were no longer statistically significant (ever users: relative risk (RR) = 1.08, 95 percent confidence interval (CI): 0.95, 1.24; use before 1976: RR = 1.11, 95 percent CI: 0.95, 1.30). Recent use (in the 3-6 years before disease diagnosis) was not associated with breast cancer rates. In age-adjusted analyses there was a suggestion of an elevation in risk with greater than or equal to 10 years of use. However, there was no evidence of a duration-response relation in this group, and the relativie risks were not statistically significant in multivariate analyses.
In the retrospective analysis, 2,426 cases were identified during 1,318,683 person-years of follow-up. There was no association between ever use of an electric blanket and breast cancer risk (multivariate RR = 1.00, 95 percent CI: 0.92, 1.09). Because of the way the question was asked, it was difficult to determine whether use between 1976 and 1991 actually occurred before or after breast cancer diagnosis. When the exposure definition was restricted to use before 1976, the date at which disease follow-up began, the age-adjusted relative risk was elevated (RR = 1.12, 95 percent CI: 1.01, 1.23) but was attenuated in multivariate analyses (RR = 1.05, 95 percent CI: 0.95, 1.16). Similarly, the evidence of a duration-response relation after controlling for breast cancer risk factors was not strong (table 3).
At the beginning of the prospective follow-up in 1992, very few women were still premenopausal (total number of women = 13,095, number of cases = 95), and the prospective analyses restricted to premenopausal women were limited by small numbers (table 4). In the retrospective analysis over half of the person-time was contributed by premenopausal women. In this analysis the age-adjusted relative risk for use before 1976 was elevated (RR = 1.23, 95 percent CI: 1.05 1.46), but as with the overall retrospective analysis this elevation was attenuated after controlling for known breast cancer risk factors (RR = 1.15, 95 percent CI: 0.98, 1.37). In the multivariate analyses, women who reported 7-14 years of exposure had an elevated risk, although, women with exposure of greater duration did not (p for trend = 0.11). The results from the prospective analysis restricted to postmenopausal women (table 5) were similar to the results for all women because the majority (76 percent) of the participants were postmenopausal in 1992. Postmenopausal hormone use and previous diagnosis of benign breast disease individually contributed most to the attenuation of the age-adjusted relative risks in the analysis of categories of total use. If the association between electric blanket use and breast cancer is real, then it is possible that benign breast disease is on the causal pathway and should not be controlled for as a confounder. However, the results obtained after controlling for all established risk factors excluding benign breast disease were similar to those reported in table 5 where benign breast disease was included.
Because the hypothesized mechanism for an effect of EMFs on breast cancer risk is through a hormonal pathway, we evaluated the possibility of effect modification by body mass index and use of postmenopausal hormones. There were no consistent patterns of risk across different categories of these factors or between the retrospective and prospective analyses.
Use of a heated waterbed represents an exposure to EMFs during the night-time hours, although it is of much lesser magnitude than that due to electric blankets because the heat source is farther from the body. When the cohort was restricted to women who did not report any waterbed use, the results for electric blanket use remained materially the same. Separate analyses of waterbed use, similar to those performed for electric blanket use, were not performed because of the limited number of cases of invasive breast cancer that occurred among women who were exposed to waterbeds and not electric blankets (12 prospective and 17 retrospective cases).
Of the cases for which estrogen receptor status was known, 80 percent of the postmenopausal cases and 72 percent of the premenopausal cases were estrogen receptor positive. As with total breast cancer, there was no evidence of an elevated rate among ever users of electric blankets in either the prospective (total number of cases = 612; multivariate RR = 1.02, 95 percent CI: 0.87, 1.21) or retrospective (total number of cases = 1,310; multivariate RR = 0.94, 95 percent CI: 0.84, 1.05) analysis. The multivariate relative risk for use before 1976 was also similar to the result from the prospective analysis of all breast cancer cases (RR = 1.04, 95 percent CI: 0.86, 1.27).
DISCUSSION
Overall, these results do not support an association between breast cancer risk and lifetime exposure to EMFs from electric blankets. For all women combined and for postmenopausal women only, there was no evidence that exposure close to diagnosis increased breast cancer in general or estrogen receptor-positive breast cancer in particular. There was a suggestion in prospective age-adjusted analyses that use at least 16 years before disease diagnosis and use of greater than or equal to 10 years might be associated with an increased risk of breast cancer. However, the relative risk was attenuated and not statistically significant after controlling for well-established breast cancer risk factors. Results from retrospective analyses were consistent with these findings.
The most convincing biologic mechanism relating exposure to EMFs and increased breast cancer risk centers on the assumption that exposure to EMFs suppresses melatonin production. There is limited experimental evidence supporting this hypothesis that shows an association of melatonin with mammary carcinogenesis (4). Melatonin is hypothesized to increase carcinogenesis through either an inverse association with estrogen (2) or direct oncostatic properties. Melatonin has been identified as a hydroxyl radical scavenger (16), and it has been observed to inhibit tumor growth in vivo and in vitro (17). If suppression of melatonin results in increased breast cancer risk due to the estrogen pathway, then recent exposure to EMFs might specifically encourage the progression of estrogen receptor-positive breast cancer tumors (18). We did not observe an association with recent use in general or specifically for estrogen receptor-positive breast cancers.
The majority of the epidemiologic studies of exposure to EMFs and breast cancer have evaluated risk in populations employed in electrical occupations (e.g., electricians, telephone linemen, communication workers, railroad workers, and power plant operators). Evidence of an elevated risk of male breast cancer, a relatively rare disease, associated with presumed occupational EMFs has been observed in four cohort studies (19-22), with measures of association ranging from 1.2 to 6.5. Two case-control studies designed to examine occupational exposure to EMFs and breast cancer in women observed an approximately 40 percent increase in the risk of mortality for electrical workers (23) and for women in occupations with potential for high exposures to EMFs (24). Only the latter study accounted for potential confounding by known breast cancer risk factors.
At least four studies have examined and discussed the relation of exposure to EMFs from residence near power lines and female breast cancer in the general population. Although Wertheimer and Leeper observed an almost threefold increase in premenopausal breast cancer risk associated with higher exposure among 58 case-control pairs (25), larger cohort studies (26, 27) and a case-control study of 1,980 cases (28) did not observe an association.
To date there have been three published evaluations of breast cancer and EMFs focusing on electric blanket use. Vena et al. (6, 7) in a case-control study obtained detailed information on electric blanket use during the 10 years prior to diagnosis, including frequency of use by season and whether the blanket was turned on throughout the night or only to warm the bed. Any use of electric blankets continuously throughout the night during the 10-year period compared with never use was associated with marginally statistically significant increases in postmenopausal breast cancer (odds ratio (OR) = 1.46, 95 percent CI: 0.96, 2.20) (6) and in premenopausal breast cancer (OR = 1.43, 95 percent CI: 0.94, 2.17) (7), controlling for age, education, age at first pregnancy, number of pregnancies, age at menarche, family history of breast cancer, body mass index, and history of benign breast disease. There was no evidence of a duration-response when exposure was defined by frequency of use during season. Using data from the postmenopausal women, Wertheimer and Leeper (29) recalculated the odds ratios, including in the reference group the women who used the blanket only to warm the bed, and obtained a statistically significant increase in risk of breast cancer (OR = 1.51, 95 percent CI: 1.03, 2.22). In response to comments (30), Vena et al. combined the results from the premenopausal and postmenopausal women; the odds ratio for all women was 1.45 (95 percent CI: 1.08, 1.95).
Gammon et al. (8) evaluated regular use of electric blankets, electric mattress pads, and heated waterbeds in a case-control study consisting of 2,170 cases of in situ or invasive breast cancer and 1,987 controls. They did not distinguish among the three different types of appliances, and regular use was not defined. The majority of the women were under the age of 45. The odds ratio for ever use, adjusted for age, center, marital status, education, race, body mass index, exogenous estrogen use (oral contraceptives and postmenopausal hormones), and alcohol consumption, was 1.01 (95 percent CI: 0.86, 1.18). The odds ratio was not elevated if the device was left on most of the time while sleeping, and the odds ratio did not increase significantly with increasing duration of use. A population-based case-control study evaluating exposure to EMFs from various sources on Cape Cod, Massachusetts, also showed no association of breast cancer risk with using an electric blanket (OR = 1.0, 95 percent CI: 0.7, 1.4) (31).
In our analyses during the equivalent time period covered by Vena’s analysis, the upper bounds of the 95 percent confidence interval were considerably below the 1.4 observed in that study. Our point estimates of the relative risk were on average higher than those observed by Gammon et al., but they were not statistically significant.
There are several limitations to our investigation associated with the assessment of exposure. Exposure was determined by self-report of electric blanket use as opposed to actual field measurements of EMFs. Characteristics of the electric blanket that can influence the intensity of exposure, such as the heating cable type and routing, the blanket size, and the thermostat design (9), were not obtained. Electric blankets designed after 1990 produce significantly lower magnetic fields than did older models (32), which could possibly affect conclusions about exposures during the last time period if women were using newly purchased blankets. The question on the 1992 questionnaire asked specifically for information on “regular use throughout the night.” Among the women who participated in the 1994 reproducibility study, the correlations of the responses to the 1992 questions with the responses to the more detailed questions asked in 1994 were good. Thus, we feel confident that we acquired information on the longest users and that we appropriately identified nonusers of electric blankets. However, details on how many hours a night, how many nights a week, or how many months a year a woman used her blanket were not obtained. Participants provided information on past use of electric blankets and heated waterbeds in broad prespecified categories, making it difficult to assess actual timing of exposure and duration of use. This limitation particularly affects the analysis of years of total use. Misclassification of exposure could obscure a modest positive association between electric blanket use and breast cancer.
Exposures to other sources of EMFs, occupational and residential, were not queried because we felt that they could not be reliably determined by a questionnaire. Household appliances, such as electric razors, hair dryers, and electric analog clocks, can have high peak magnetic fields (32). However, the time-weighted contribution to whole-body exposure from these sources is negligible compared with usual background exposures, because they are used close to a small portion of the body for only a short period of time (33). The magnitude of exposure from power lines is highly dependent on the duration of the exposure and on the distance from the source. A survey of nurses’ occupational exposure to EMFs in a local hospital, performed by one of us (Tolbert, unpublished data), indicated that only two relatively unusual sources provided unusually high exposures: phototherapy lights used in neonatal intensive care and dialysis machines. Electric blanket use should supply an increase in exposure sufficiently greater than background, and it is unlikely that use of an electric blanket would be correlated with exposures to other sources of EMFs.
Although there is the possibility for recall bias and selection bias in the retrospective analysis, we expect it to be minimal. The hypothesis that electric blanket use may be associated with an increased risk of breast cancer was not well known in the general population in 1992 when the question was asked. Selection bias could have been introduced through the exclusion of cases who died before the assessment of electric blanket exposure in 1992. Seven hundred deaths (13 percent from invasive breast cancer cases diagnosed after 1976) were attributed to breast cancer during the period of follow-up. If heavy electric blanket use was associated with the most aggressive cancers, then the association would be underestimated. However, there is no evidence of such an association in previous studies, and the retrospective and prospective results in this study were similar.
The strength of our study lies in the prospective follow-up component and the large numbers included in the retrospective analysis. The prospective analysis eliminated potential concerns about recall bias and change of lifestyle due to diagnosis. In addition, because of the extensive characterization and thorough follow-up of this population, we were able to adequately control for known risk factors for breast cancer and potential confounders of the EMFs-breast cancer relation. The ability to control confounding is of particular importance with risks of relatively low magnitude, such as those hypothesized for EMFs.
In conclusion, our findings did not support an important association between exposure to EMFs from lifetime electric blanket use and breast cancer. Our findings were generally null, although the confidence intervals around our risk estimates did not exclude a small excess risk in those with the longest duration of exposure many years before diagnosis. Additional prospective follow-up designed to specify timing and duration of use will assist in the evaluation of this relation.
This research was supported by National Cancer Institute grants CA/ES62984 and CA40356. Dr. Laden was supported in part by an Institutional National Research Service Award in Cancer Epidemiology, 5T32CA09001. Dr. Hunter was also partially supported by American Cancer Society Faculty Research Award FRA-455.
The authors thank the registered nurses who participated in this study, Gary Chase, Karen Corsano, Albert Liu, and Barbara Egan.